fix(translation): decode OpenRouter cache write usage

[nachiketb-nvidia]

Summary

• decode OpenRouter's current prompt_tokens_details.cache_write_tokens field in buffered OpenAI Chat responses
• decode the same field from streaming usage chunks
• retain cache_creation_tokens as a compatibility alias

Root cause

OpenRouter reports cache writes as cache_write_tokens, while the OpenAI Chat codecs only recognized cache_creation_tokens. Cache-write tokens were therefore left in the uncached input count and omitted from normalized cache-creation usage.

Validation

• cargo fmt --all -- --check
• cargo test -p switchyard-translation
• cargo clippy -p switchyard-translation --all-targets -- -D warnings

Summary by CodeRabbit

• Bug Fixes

  • Improved token usage reporting for OpenAI-compatible responses and streaming events.
  • Cache-write token counts are now recognized consistently, including support for legacy cache-creation fields.
  • Corrected input, cached, cache-creation, and output token totals in translated usage data.

• Tests

  • Added coverage confirming equivalent handling of both cache-write field formats.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 004d2a67-e264-4e8c-9ef7-c3c8c0bf5f20

📥 Commits

Reviewing files that changed from the base of the PR and between 5349735 and dd02faa.

📒 Files selected for processing (4)

• crates/switchyard-translation/src/codecs/openai_chat/buffered.rs
• crates/switchyard-translation/src/codecs/openai_chat/stream.rs
• crates/switchyard-translation/tests/response_translation.rs
• crates/switchyard-translation/tests/stream_translation.rs

---

Walkthrough

OpenAI Chat usage normalization now prefers cache_write_tokens and falls back to cache_creation_tokens in buffered and streaming paths. Tests cover cache read, cache creation, input, and output token mappings for both field names.

Changes

OpenAI cache usage

Layer / File(s)	Summary
Cache-write token normalization crates/switchyard-translation/src/codecs/openai_chat/buffered.rs, crates/switchyard-translation/src/codecs/openai_chat/stream.rs, crates/switchyard-translation/tests/response_translation.rs, crates/switchyard-translation/tests/stream_translation.rs	Buffered and streaming usage decoding prefer cache_write_tokens, fall back to cache_creation_tokens, and test normalized cache read, cache creation, input, and output token values.

Estimated code review effort: 2 (Simple) | ~10 minutes

Poem

A bunny found tokens tucked under a leaf,
Cache-write carrots, fresh and brief.
Legacy crumbs still lead the way,
Read and output count the day.
Hop, hop—usage blooms anew!

---

_{Comment @coderabbitai help to get the list of available commands.}